What this connector does
The PitCrew MCP Connector exposes your PitCrew workspace to MCP clients, scoped to running your published agents. Through a single conversation in your MCP client you can:
- Trigger a run of any published PitCrew agent with inputs you provide in chat or upload via an interactive form
- Attach documents to runs, either freshly uploaded or pulled from a collection you already own
- Submit compliance validations against policies in your organisation’s policy library
- Read back results, extracted variables, citations, and compliance decisions directly in the conversation
- Send the final report through your connected Gmail account
The connector does not provide a conversational Builder for designing new agents from scratch. Agents must be published in your PitCrew workspace via the PitCrew web app before this connector can run them.
The connector speaks Model Context Protocol (MCP) over Streamable HTTP, and uses OAuth 2.1 (PKCE) to authenticate against your PitCrew account.
Connect your MCP client to PitCrew
Add the PitCrew connector in your MCP client using the URL below. On first use, the client shows a sign-in prompt that opens the PitCrew login page in your browser. After you authenticate, the connector is linked to your PitCrew account and will not prompt again until the token expires.
send_email), you must connect those services in your PitCrew account before the related tool calls will succeed.
All 29 tools, grouped by domain
Every tool is annotated so the MCP client can decide whether to auto-run it (read) or prompt for your approval (destructive). The first label on each row is the user-facing title (shown when the tool is invoked). The grey chip underneath is the internal tool name used in logs and MCP Inspector.
Agents
Discover, run, and track published PitCrew agents.
| Tool | Hint | What it does |
|---|---|---|
| List agents list_agents | read | List the published agents available in your organization. |
| Start agent run start_agent_run | destructive | Start a run of a published agent with inputs you already have, optionally with files or collection documents. |
| Open agent input form open_agent_input_form | read | Open an interactive input form (with file drop zones) so you can fill in run inputs; the run starts when you submit the form. |
| Get agent run status get_run_status | read | Return the current progress of a run, or its final output when complete. |
| Publish agent publish_agent | destructive | Persist a complete agent spec to your workspace. Used by tooling that produces agent specs programmatically; not intended as a hand-written user flow. |
Collections
Group related documents (PDFs, CSVs, XLSXs) under one named bucket so they can be reused across many agent runs without re-uploading.
| Tool | Hint | What it does |
|---|---|---|
| Create collection create_collection | destructive | Create a new named collection with a domain (general, financial, contracts, identity, tax, or legal). |
| List collections list_collections | read | List collections visible to your organization. |
| Get collection details get_collection | read | Fetch a single collection and its documents (document IDs, filenames, ingest status). |
| Get collection ingest status get_collection_status | read | Lightweight ingest-status digest, suitable for checking back after an upload. |
| Upload documents to collection upload_documents | destructive | Upload one or more documents to a collection (base64-encoded, up to 10 MB per file). |
Validation
Submit compliance validations against your policy library and retrieve the resulting report.
| Tool | Hint | What it does |
|---|---|---|
| Run compliance validation submit_validation_async | destructive | Submit a compliance validation job against an entire policy. |
| Run targeted compliance validation submit_targeted_validation_async | destructive | Submit a validation job against specific policy citations (chunks) rather than the full policy. |
| Check validation status check_validation_status | read | Return the current status of a validation job. |
| Get validation results get_validation_results | read | Retrieve the complete validation report (satisfied rules, violations, extracted variables, compliance scores). |
| Get variable extraction reasoning get_variable_reasoning | read | Return the reasoning behind each extracted variable in a completed validation. |
Policies
Browse policies in your organization’s policy library and inspect their rules and citations.
| Tool | Hint | What it does |
|---|---|---|
| List policies list_policies | read | List the policies in your organization’s policy library. |
| Get policy rules and variables get_rules_and_variables | read | Return every rule and variable in a specific policy, with source citations. |
| Get policy citations get_policy_citations | read | Return only the citation mapping for a policy. Lighter payload than full rules. |
Clients
Client-management tools for RIA (Registered Investment Advisor) workflows.
| Tool | Hint | What it does |
|---|---|---|
| List clients get_all_clients | read | List all clients in your CRM. |
| Get client details get_client_by_id | read | Fetch complete details for a specific client, including uploaded documents. |
| List client source documents get_client_sources | read | List the original source documents uploaded for a client. |
| Get client extracted fields get_extracted_fields | read | Return all extracted fields on a client, with confidence and source metadata. |
| Update client field update_or_create_field | destructive | Update an existing field, or create a new one if it does not exist. |
| Delete all client fields delete_client_fields | destructive | Delete all extracted fields for a client. Bulk reset operation. |
Documents and search
Fetch source text and run semantic searches against the document corpus.
| Tool | Hint | What it does |
|---|---|---|
| Get RAG source text get_rag_source_text | read | Return the full text of a RAG chunk by its source ID. |
| Search citations get_relevant_citations | read | Semantic search for relevant citations across your document corpus. |
Integrations
Tools that reach outside PitCrew. The amber hint indicates an external destination.
| Tool | Hint | What it does |
|---|---|---|
| Send email send_email | destructive · external | Send an email through your connected Gmail account. Requires Gmail connected in your PitCrew account. |
Utility
Small sanity-check tools used for connectivity diagnostics.
| Tool | Hint | What it does |
|---|---|---|
| Echo text echo | read | Echo the input text back. |
| Add numbers add_numbers | read | Add two numbers. |
Real UI, inside the chat
For agents whose inputs are not easy to describe in chat (notably file inputs), the connector renders a real UI form inline in the conversation using the MCP Apps UI extension (io.modelcontextprotocol/ui).
The form includes drag-and-drop zones for file inputs, text fields for scalar inputs, and a Run button. When you submit the form, the run starts and a confirmation message with the run_id is posted back into the chat.
Inputs can be mixed in a single run: files uploaded fresh via the form, documents pulled from a collection, and scalar fields filled in chat.
Where your data goes
Documents and inputs you process through the connector flow to your PitCrew workspace. Your organisation’s data retention, access controls, and compliance policies apply.
Where data goes
Documents, extracted fields, run inputs, and run outputs are stored in your PitCrew workspace on PitCrew infrastructure.
Third-party services
Tools that reach outside PitCrew (for example, send_email to Gmail) call only accounts you have explicitly connected in your PitCrew settings. The connector itself does not initiate connections to third-party services on your behalf.
Conversation data
The connector does not read the client’s memory, chat history, conversation summaries, or user files. Each tool call carries only the arguments the tool needs.
Retention
Governed by your PitCrew account settings and organisation policy. See the PitCrew Privacy Policy for defaults and the PitCrew Security Overview for the controls PitCrew applies.
Reasonable defaults
The connector proxies calls to the PitCrew backend, which applies per-organisation rate limits. Normal interactive use does not approach these limits. High-volume automated use may hit them; contact support@gopitcrew.com for quota requests.
Common issues
“No Gmail integration found” when calling send_email
Visit your PitCrew account at gopitcrew.com and connect Gmail under integrations, then retry.
File too large
Each file sent via upload_documents or an input form is capped at 10 MB. For larger files, split the upload or contact support for a quota adjustment.
Form submits but nothing happens
Confirm the connector is connected in your client’s settings panel. If the issue persists, capture the time of the attempt and the agent ID, then contact support.
OAuth prompt loops
Sign out of PitCrew in your browser, disconnect the connector in your client, and reconnect. If the loop continues, contact support with your account email.
Get help
Email support@gopitcrew.com with your account email, the time of the issue, and the agent or run ID when relevant.